anthropic has announced that it will not release its latest frontier model, Claude Mythos Preview, to the General public. This strategic decision follows the model's alArming discovery of thousands of high-severity, zero-day vulnerabilities across critical global software infrastructure.
To mitigate the risks of these capabilities falling into the wrong hands, Anthropic has launched Project Glasswing, a major industry coalition dedicated to proACTive software security.
🚨 The Security Paradox: Power vs. Risk
The decision to restrict public access stems from the model's unprecedented ability to identify security flAWS. claude Mythos Preview has successfully uncovered thousands of bugs, including zero-day vulnerabilities that have evaded human detection for decades.
"AI cyber capabilities at this level will proliferate over the coming months, and not every actor who gets access to them will be focused on defense. That's the gap Glasswing is built to close."
— Daniela Amodei, President at Anthropic
If released without safeguards, these advanced capabilities could enable bad actors to mount sophisticated, large-scale cyberattacks.
🐛 Critical Vulnerabilities Discovered
The efficacy of Claude Mythos Preview was proven through its analysis of hardened systems. Key findings include:
OpenBSD: A nearly 30-year-old vulnerability was found in this security-hardened OS, allowing attackers to rEMOtely crash machines.
Linux Kernel: Flaws were exposed that could allow an attacker to take complete control of the world's servers.
FFmpeg: Critical bugs were identified in the video encoding/decoding software used globally.
🤝 Project Glasswing: A Coalition for Defense
To address these risks, Anthropic formed Project Glasswing, a collaborative effort to secure critical software using AI. This coalition includes industry giants such as:
Tech & Cloud: AWS, Apple, Google, Microsoft, NVIDIA, Broadcom, Cisco
Security: CrowdStrike, Palo Alto Networks
Finance & Open Source: JPMorgan Chase, The LinUX Foundation
George Kurtz, CEO of CrowdStrike, noted, "AI is creating the largest security demand driver since enterprises moved to the cloud... Mythos may be the most capable frontier model yet."
💰 Commitment to Responsible Innovation
Anthropic is backing this initiative with significant financial resources to ensure the responsible scaling of AI:
$100 Million: Committed by Anthropic in model usage credits for Project Glasswing participants.
$4 Million: Donated to open-source security organizations to support vulnerability patching.
🚀 The Future of AI-Driven Security
By providing controlled access to organizations like Cisco, Anthropic aims to fix flaws faster than they can be exploited.
"AI capabilities have crossed a threshold that fundamentally changes the urgency required to protect critical infrastructure," said Anthony Grieco of Cisco. "Project Glasswing demonstrates that securing advanced AI systems requires widespread industrial cooperation."
Comments & Questions (0)
No comments yet
Be the first to comment!